ezcut - AI-Powered Video Story Extraction Platform

TRY OUR LINK IT ACTUALLY WORKS (takes a while though)

Inspiration

Honestly we were just tired of spending hours going through footage trying to find the good parts. Like, streamers and vloggers have SO much content but picking out the best moments for shorts is such a pain. There's automated tools but they just make random cuts without thinking about the actual story you're trying to tell. We thought there had to be a better way.

What it does

ezcut basically lets you upload a bunch of videos and then tells you what stories it thinks you can make from them. But here's the thing - you can actually customize what story you want instead of just getting some random AI-generated mess.

So you upload your files, we process them with transcript stuff and visual analysis (CLIP and FAISS - fancy AI things), then our NLP figures out potential narratives. You pick what you want to focus on (like we showed with Jonathan's keynote) and boom - it makes you a shorter video that actually tells the story you wanted.

How we built it

Ok so we threw together a bunch of different technologies:

Simple upload UI (nothing fancy, just works)
Video processing with transcripts and timestamps
CLIP for understanding what's actually happening in the frames
FAISS for searching through all the video content efficiently
Some NLP magic to figure out story elements
Backend that somehow makes all this work together

Basically we're using AI to understand both what people are saying AND what's visually happening, then matching that to the story the creator actually wants to tell.

Challenges we ran into

Processing videos takes FOREVER especially with all the AI analysis
Getting the story extraction to actually make sense was way harder than expected
CLIP and FAISS are powerful but figuring out how to use them together efficiently was a nightmare
Making the UI simple when there's so much complexity happening behind the scenes
Video quality kept getting messed up during processing and we had to fix that
Honestly just getting everything to work together without crashing was a challenge

Accomplishments that we're proud of

We actually got it working! Like, you can upload videos and get back something that makes sense story-wise. The demo worked (thank god) and people seemed to get what we were trying to do. Also managed to integrate all these different AI technologies without everything falling apart, which felt pretty good.

The story-based approach is definitely what sets us apart from other tools, and seeing it work on real content was really satisfying.

What we learned

Video processing is way more complicated than we thought
Creators really do want control over their narrative, not just random automated cuts
AI is great but you still need to think about the user experience
Performance optimization is super important when you're dealing with large video files
Sometimes the simplest UI is the hardest to build because you're hiding so much complexity